CN100375398C

CN100375398C - Colored interference identification

Info

Publication number: CN100375398C
Application number: CNB038120127A
Authority: CN
Inventors: 何首生
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2002-05-27
Filing date: 2003-05-16
Publication date: 2008-03-12
Anticipated expiration: 2023-05-16
Also published as: EP1508201A1; WO2003100997A8; DE60307270T2; CN1656698A; WO2003100997A1; DE60307270D1; US7319846B2; ATE335312T1; EP1508201B1; AU2003240664A1; US20050227663A1

Abstract

A method and apparatus for receiving a communication signal r(t) subject to noise n(t) over a communication channel (10) is disclosed. The method comprises the steps of: receiving (100) the communication signal r(t) comprising the noise n(t), estimating (103) the amount of noise n(t) in the communication signal r(t), and determining (105) the hue of the estimated amount noise n(t), wherein the received communication signal r(t) is passed through a whitening filter if the hue of the noise n(t) is greater than a predetermined threshold level.

Description

Colored interference identification

Technical Field

The present invention relates generally to a communication method and apparatus, and more particularly, to a method and apparatus for determining a hue of a noise component introduced when a communication signal is transmitted in a wireless communication system.

Background

In recent years, many different wireless communication systems have been used to provide voice and data services to users. One well-known digital cellular mobile telephone standard is GSM (global system for mobile communications), which has covered a large portion of the world. In the following, the GSM system will be discussed as a basic example, but the following description is essentially applicable to other mobile phone standards, such as D-AMPS (advanced digital mobile phone system) or PDC (pacific cellular digital system).

The performance of any communication system available today is affected by noise from a variety of sources. It is well known that thermal noise can occur in any electronic system, for example due to random motion of electrons in the system as charge carriers. The thermal motion of the electrons creates what we often speak of "white noise".

The term "white" here comes from the fact that the energy of the noise is evenly distributed over the spectrum, i.e. over a longer time, the energy is distributed in the range of 0-10KHz and in the range of 100-110 KHz.

In addition to white noise, colored noise is also present in most electronic systems today. Colored noise can come from a variety of sources, such as metal oxide junctions in Field Effect Transistors (FETs), which produce noise known as pink noise or flicker noise. As we will discuss below, the interference of the shared channel and the adjacent channels in the communication system can also be colored noise sources.

Colored noise is characterized by an uneven distribution of noise energy across the spectrum. For example, the distribution of pink noise over each octave is the same, i.e., the energy distribution between 10KHz and 20KHz is the same as the energy distribution between 100KHz and 200 KHz. Many other forms of colored noise are defined and described throughout the text, but they all exhibit the basic feature of uneven noise energy distribution.

In wireless communication systems, the background noise in the communication channel is typically white noise. However, the performance of these systems is not limited only by background noise, but rather by other users from the system. It is well known in the art that a multiple access scheme defines how different communications take place simultaneously between different mobile stations in different cells sharing a radio frequency band. In the GSM system, the multiple access scheme is a hybrid scheme using FDMA (frequency division multiple access) and TDMA (time division multiple access).

More specifically, in the GSM system, 124 carrier frequencies having a bandwidth of 200KHZ constitute a 25HMZ band using the FDMA scheme. And each of the 124 carrier frequencies is further divided in time using a TDMA scheme. This scheme divides a 200KHZ wide radio channel into 8 bursts. A user is assigned a burst for communication.

There are generally two types of interference for narrow band TDMA systems such as GSM. Common channel interference is generated for users using the same carrier, while adjacent carrier users generate adjacent channel interference. As mentioned earlier, noise due to common channel and adjacent channel interference appears in the colored spectrum and the noise energy distribution is not uniform. Also, the receiver filter (narrower than the Nyquist bandwidth), which is typically at the receiver input, may also make the background noise appear as colored noise.

Colored noise significantly degrades the performance of the Maximum Likelihood Sequence Estimation (MLSE) equalizer of the receiver, which is optimal only under the assumption that the current noise is additive white gaussian noise (WGAN).

In order to combat the performance degradation caused by colored noise, a "whitening filter" may be introduced before the equalizer. In addition, an unbiased channel estimate ("blue, best linear unbiased estimate") may also be needed. Both whitening filter settings and unbiased channel estimation require knowledge of the noise characteristics, which can be obtained by an initial estimate of the autocorrelation of the noise, and the signal information (e.g., in a GSM system, through a training sequence).

As described above, the whitening filter and unbiased channel estimation significantly improve equalizer performance when the noise exhibits strong colored characteristics (i.e., the hue of the noise is very high), such as when a strong adjacent channel interference is present. However, when the noise is close to white noise, the whitening filter and unbiased channel estimation may result in performance degradation, which is very significant in certain situations, such as in a mountainous area environment. This is because a limited training sequence length is accompanied by an under-estimation of the noise characteristics.

At certain levels of noise hue, the benefits of whitening will be outweighed by the impairment due to the lack of noise estimation. Further, the whitening filter and unbiased channel estimation will increase the computational burden on the signal processing unit in the system.

WO 0139448 A1 discloses a system for whitening an interference signal in a communication signal by means of a filter whose coefficients are adaptively established by means of information in each burst of the received signal. In one embodiment disclosed in WO 0139448 A1, the received signal is processed through a whitening filter having M +1 taps, where M is a selected integer. The coefficients of this whitening filter are based on an M-th order linear predictor of signal interference. Alternatively, the coefficients may be based on autocorrelation of the signal interference. The process for performing signal whitening places high demands on the processor performing the computation even if there is no or little colored noise signal in the signal, because the whitening process is performed independently of the current noise tone.

US 5031195 discloses an adaptive modem receiver comprising an adaptive Whitening Matched Filter (WMF). The WMF includes an adaptive linear equalizer and an adaptive linear predictor. The coefficients of the predictor are updated such that the noise at the input of the subsequent sequence decoder is whitened regardless of whether the additive noise from the communication channel path is correlated. There is no way to provide any method for reducing the computational burden even if the noise is white noise or very lightly tonal.

US 5283811 discloses a decision feedback equalizer that enhances the performance of the receiver when the transmitted signal is affected by multipath propagation, which consequently causes propagation delays and intersymbol interference. Where the delay spread due to multipath propagation is weak (i.e., less than one third of the symbol duration), the equalizer may be switched out of the circuit. However, no device is disclosed in US 5283811 as to how the receiver performance is improved in accordance with the present colored noise caused, for example, by adjacent channel interference.

US 2002/02/34720 discloses a method for removing colored noise from a signal in a wireless communication system. A signal including noise is received and an amount of noise in the signal is estimated. However, this method also discloses that the above-mentioned problem is encountered, i.e., the system performance is degraded because the system receives a signal containing white noise, adjacent channel and common channel interference.

Disclosure of Invention

The present invention seeks to provide a method to improve the performance of a communication device when receiving signals affected by noise, for example due to adjacent channel interference and common channel interference.

This object has been achieved by a method for receiving a communication signal r (t) affected by noise n (t) in a communication channel, comprising the steps of: 1) receiving (100) a signal r (t) comprising noise n (t), 2) estimating (103) the energy of the noise n (t) in the communication signal r (t), 3) determining (105) the hue of the estimated noise n (t), passing the received communication signal through a whitening filter if the hue of the noise n (t) is above a predetermined threshold, or bypassing the whitening filter if the hue of the noise n (t) is below the predetermined threshold.

According to a preferred embodiment, this method is performed by a communication device comprising: receiving circuitry (30, 40, 50) for receiving a signal r (t) affected by noise n (t) in a communication channel, and a signal processing unit (60) adapted to: 1) Estimating (103) the energy of noise n (t) in the communication signal r (t), 2) determining (105) the hue of the estimated noise n (t), switching in (107) a whitening filter (80) if the hue of the noise n (t) is above a predetermined threshold, or bypassing the whitening filter if the hue of the noise n (t) is below the predetermined threshold.

Other objects, features and advantages of the present invention will become more apparent from the following detailed disclosure of the preferred embodiments:

drawings

Preferred embodiments of the present invention will now be described with reference to the accompanying drawings, in which:

FIG. 1 is a schematic block diagram illustrating different processing blocks in accordance with a preferred embodiment;

FIG. 2a is a diagram illustrating the result of calculating the autocorrelation of an information sequence r (m);

FIG. 2b is a diagram illustrating the result of calculating the autocorrelation of white noise;

FIG. 2c is a diagram illustrating the result of calculating the autocorrelation of colored noise;

fig. 3 is a schematic flow chart illustrating the steps of determining the hue of a noise disturbance in a signal according to a preferred embodiment.

Detailed Description

Fig. 1 provides an overall depiction of a receiver in a communication system in accordance with a preferred embodiment of the present invention. The information signal s (t) is transmitted in the form of radio waves in the communication channel 10. However, the medium carrying the information is of secondary importance to the functioning of the invention, and the information may equally well be transmitted via optical, cable or other suitable communication medium. For reasons of simplicity, however, only communication via radio waves will be discussed herein.

In fig. 1, the signal s (t) modulates a high frequency downlink carrier, which in the case of GSM communications is in the range 935-960MHZ, before being transmitted by the transmitter. The output of the transmitter is thus a high frequency signal (i.e. a carrier envelope modulated by the signal s (t)) suitable for transmission.

Regardless of which communication medium is selected for transmitting the information signal s (t), the signal s (t) will be altered, wherein interference n (t) associated with the characteristics of the channel 10 will be introduced during transmission over the actual communication channel 10. As mentioned earlier, interference from different sources occurs, with common channel interference and adjacent channel interference being dominant.

The high frequency signal is received at a high frequency circuit 30, which high frequency circuit 30 in the preferred embodiment operates according to the homodyne principle, whereby the received information signal r (t) is extracted from the received high frequency signal by mixing the signal of the local crystal oscillator 31 with the received high frequency signal. Generally with the above mentioned signal modulation, demodulation according to the homodyne and heterodyne principles is well known in the art and can be easily found in the literature. However, any other suitable demodulation scheme is possible within the scope of the invention.

After removing the high frequency carrier, the received baseband signal r (t) is transferred to an analog-to-digital converter (ADC) 40 to convert the analog signal r (t) into a time-discrete digital signal r (m). The sampled and converted signal r (m), after filtering in the low-pass filter 50, is then received by a signal processing unit 60, which signal processing unit 60 is in the form of a DSP (digital signal processor) in the preferred embodiment, which DSP performs the steps disclosed below by executing executable program code stored in a memory 61. However, the signal processing unit may also be implemented in the form of, for example, an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).

The signal processing unit performs a first initial channel estimation after burst synchronization based on known signal information in the received signal r (m), i.e. a training sequence in the GSM/EDGE case. The received signal r (m) will be compared to the desired symbol sequence to determine the noise samples n (m) according to the following formula:

the embedded training sequence is generally of a smaller length. This means that it is difficult to determine the characteristics of the noise by analyzing the sequence in the time domain. In this case, the autocorrelation calculation of the noise n (t) is a very powerful tool that can be used to obtain information about the noise spectrum.

Typically, the noise autocorrelation estimate (assuming a mean of zero) is calculated from the estimated noise samples by the following equation.

Where Ns is the number of estimated noise samples, () ^* Representing a complex conjugate. The autocorrelation of the noise is usually a complex, conjugate symmetric sequence, so that a negative index sequence can be obtained by a positive index sequence, ρ _-k ＝ρ ^* . In the above equation, ρ ₀ Always one real unit.

Fig. 2a illustrates the result of calculating the autocorrelation of the information sequence r (m). As can be seen from fig. 2a, the result of the autocorrelation is a vector, here represented as a graph, centered on the Y-axis (i.e., zero lag), which decays towards zero as the delay increases (or increases in advance once the subsequent signal values are known). Thus, the information sequence shows a high degree of autocorrelation in adjacent and near-adjacent samples.

If no information signal is present, i.e. the signal consists of noise n (m) only, as seen in fig. 2b, the autocorrelation for each delay is almost 0 (except for the delay =0, which is always defined by 1) and there is no significant peak as found in fig. 2 a.

Fig. 2c shows the autocorrelation of the noise in the case where the noise is not white, not evenly distributed over the spectrum, and is colored in the sense that the noise energy is higher in some parts of the spectrum. Such noise may occur, for example, in channels with strong adjacent channel interference. As can be seen from fig. 2c, the colored noise shows a certain degree of autocorrelation, since the output of the autocorrelation calculation is not zero and will decay for non-zero delays. A more tonal noise (e.g., a higher tonal value) may result in an autocorrelation calculation output having a non-zero delay value that is greater than the weak tone signal interference.

The signal processing unit 60 then determines the center of gravity of the autocorrelation function according to the following formula:

note that in the preferred embodiment, this formula uses k +1 instead of k in the numerator to preserve the weighting of the first and most important element of the noise autocorrelation function. However, other functions for determining the center of gravity may be used.

The stronger the hue of the noise, the larger the value of the center of gravity calculation result obtained. This is a corollary based on the fact that: a noise signal with a strong hue will result in an autocorrelation with a high non-zero delay value and the center of gravity is calculated for only one end of the autocorrelation, i.e. only the autocorrelation values to the right of the Y-axis are considered.

Therefore, the hue of the noise can be determined by a single variable σ. The threshold may be set in terms of N and s to turn on/off the whitening filter/unbiased estimation function. In a practical case, for example, for GSM/EDGE with a training sequence of 26 symbols in a normal burst,

the threshold may be set experimentally as:

σ _T ＝1+s

note that in this example, the threshold is not proportional to the over-sampling rate.

If σ is smaller than σ _T The noise is considered white and both the whitening filter and the non-bias estimation (BLUE) 80 are bypassed by the colored noise identification module.

Instead, a much simpler least squares estimate 70 may be used for the channel impulse response. This reduces the computational power requirements of the signal processing unit 60, which in turn makes it possible to reduce the system clock of the signal processing unit 60 by using, for example, PLL technology. It is known that reducing the clock frequency of an electronic system also reduces the power consumption of the system. Thus, for a given battery capacity, the system will be able to operate for a longer period of time.

On the other hand, if σ is greater than the threshold σ _T Then the noise is not white and the unbiased estimation and whitening filter 80 will be introduced into the system before the equalizer 90.

Fig. 3 illustrates a flow chart for determining the hue of the noise n (m). The process begins at step 100 by receiving a baseband signal corresponding to a training sequence. As described above, it is understood from the context that the information signal r (t) is generally obtained by demodulating a high frequency signal. Regardless of the transmission process (i.e., the high frequency modem process performed when transmitting a signal) prior to reception, the baseband signal received over the channel includes both the desired training sequence s (t) and the noise signal n (t).

Before further processing, the signal is converted in step 101 by an analog-to-digital converter (ADC) into a time-discrete digital signal r (m). The ADC is typically of the oversampling type (i.e. the signal is sampled at a frequency twice higher than its highest frequency component), but may also be of the Nyqvist type (i.e. the signal is sampled at a frequency twice as high as its highest frequency).

The a/D converted signal r (m) will be low pass filtered in step 102 before being transmitted to the signal processing unit 60, where a first estimation of the noise energy n (m) is performed in step 103 in the signal processing unit 60. As mentioned above, this is possible because the known signal sequence (i.e. the training sequence in the GSM system) is found in the received signal r (m). Since the channel characteristics distort the training sequence (i.e., add noise such as multipath interference and additive interference), it is possible to determine the noise by comparing the received signal r (m) with the known training signal.

The autocorrelation of the estimated noise signal n (m) is calculated in step 104. The autocorrelation may exhibit the frequency characteristics of noise even if the signal length is short. Most of the Digital Signal Processors (DSPs) available today are adapted to perform autocorrelation calculations efficiently, which means that autocorrelation calculations are not a major burden from a processing point of view.

In step 105, the center of gravity σ of the autocorrelation noise is calculated according to equation 3. If the hue of the noise n (m) is low, the center of gravity is very close to where the delay equals 0, since the energy is very small at delay < >0, as can be seen in fig. 2 b. However if the hue increases due to e.g. adjacent channel interference, the center of gravity will be pushed away from the Y-axis, as can be seen from fig. 2 c.

The signal processing unit 60 determines in step 106 whether the center of gravity σ is larger than a preset threshold σ _T Here σ _T Based on empirical values on a mathematical level.

If center of gravity σ is greater than threshold σ _T The signal processing unit 60 activates the whitening filter 80 and the unbiased channel estimation in step 107 by the colored interference identification module 100. It should be appreciated that the whitening filter/unbiased channel estimation (80) can be performed by the signal processing unit (60) alone, by a separate DSP, by fixed logic such as an FPGA (field programmable gate array), or by an ASIC (application specific integrated circuit). The whitening filter can then provide a lower-tone signal to the equalizer (or demodulator) (90), which in turn increases equalizer performance since the equalizer assumes that the interference introduced by the channel is white noise.

However, if the center of gravity σ is below the threshold σ _T Then the whitening filter will not be activated since this noise is considered to be white noise. As mentioned above, applying a whitening filter to a signal containing white noise not only increases the computational burden, but also reduces the efficiency of the decoding process in the equalizer (90) in most cases. The whitening filter and the unbiased channel estimate may be bypassed.

The invention has been described above with reference to preferred embodiments. However, other embodiments than the ones disclosed herein are possible within the scope of the invention, which is defined in the appended independent claims.

Claims

1. A method for receiving a communication signal r (t) subject to noise n (t) over a communication channel, comprising the steps of:

receiving (100) a communication signal r (t) comprising noise n (t),

estimating (103) an amount of noise n (t) in the communication signal r (t), the amount being expressed in terms of energy,

determining (105) the hue of the noise n (t) estimate,

comparing (106) the hue of the estimate of noise n (t) with a predetermined threshold, an

Passing the received communication signal r (t) through a whitening filter (80) if the hue of the noise n (t) is greater than a predetermined threshold, or

Bypassing the whitening filter (80) if the hue of the noise n (t) is less than the predetermined threshold.

2. The method of claim 1, wherein the energy of the noise n (t) is estimated by comparing the received signal r (t) with known signal information.

3. The method according to claim 1 or 2, wherein the frequency characteristic of the noise n (t) is determined (104) by performing an autocorrelation of the noise n (t).

4. The method of claim 3, wherein the noise hue is determined by a single variable.

5. The method according to claim 4, wherein the hue of the noise n (t) is determined (105) by determining the center of gravity σ of the noise autocorrelation sequence.

6. The method of claim 4 or 5, wherein the hue of the noise is determined according to the following formula:

where Ns is the number of noise samples estimated, ρ is the noise autocorrelation estimate, and k is the index of the noise sequence.

7. The method of claim 1, wherein the threshold value engages an unbiased channel estimation if the hue of the noise n (t) is greater than a predetermined threshold value.

8. A communication device, comprising:

-receiving circuitry (30, 40, 50) for receiving a communication signal r (t) affected by noise n (t) over a communication channel;

a signal processing unit (60) adapted to:

estimating (103) an amount of noise n (t) in the communication signal r (t), the amount being expressed in terms of energy;

determining (105) a hue of the noise n (t) estimate;

comparing (106) the hue of the noise n (t) estimate to a predetermined threshold;

9. The apparatus according to claim 8, wherein the signal processing unit (60) is adapted to estimate (103) the energy of the noise n (t) by comparing the received signal r (t) with known signal information.

10. The device according to claim 8 or 9, wherein the signal processing unit (60) is adapted to determine (104) a frequency characteristic of the noise n (t) by performing an autocorrelation of the noise n (t).

11. The device according to claim 10, wherein the signal processing unit (60) is adapted to determine the noise hue by a single variable.

12. The device according to claim 11, wherein the signal processing unit (60) is adapted to determine (105) the hue of the noise n (t) by determining a center of gravity σ of the noise autocorrelation sequence.

13. The apparatus according to claim 11 or 12, wherein the signal processing unit (60) is adapted to determine the hue of the noise according to the following formula:

14. The apparatus according to claim 8, wherein the signal processing unit (60) is adapted to engage the unbiased channel estimation when the hue of the noise n (t) is larger than a predetermined threshold.

15. The device according to any of claims 8,9, 14, wherein the signal processing unit (60) is a Digital Signal Processor (DSP).